Google Readies Gemini 4, Replacing Gemini 3.5 Pro Entirely
Google DeepMind confirms its next flagship AI model has entered post-training, bypassing Gemini 3.5 Pro entirely with an anticipated release target as early as October.
Key takeaways
- Google has confirmed that Gemini 4 is in post-training and will completely replace the unreleased Gemini 3.5 Pro.
- Google DeepMind leadership hopes to release Gemini 4 much earlier than the end of 2026, with an October release expected.
- Gemini 4 Pro is powered by the 'Barium B' architectural checkpoint, focusing on software engineering, 3D rendering, and real-time data analysis.
- The model is currently undergoing internal testing to power Google's Antigravity platform.
Google DeepMind is preparing to roll out its next flagship artificial intelligence model, Gemini 4, which will replace Gemini 3.5 Pro entirely across its offerings. Speaking at The Information's AI Agenda Live Summit in late September 2026, Google DeepMind head Koray Kavukcuoglu confirmed that Gemini 4 has entered post-training, the development phase where guardrails and safety testing are applied to ensure reliable behavior prior to a public debut, according to 9to5Google.
The move marks a key shift in Google's frontier model strategy after months of speculation regarding its high-end roadmap. While Google had teased Gemini 3.5 Pro during the summer after originally planning an announcement around its May 2026 developer conference, the company chose to take a step back according to 9to5Google and pivot focus entirely to Gemini 4, as reported by Geeky Gadgets.
Pivoting from Gemini 3.5 Pro to Gemini 4
Google's primary flagship model line has seen few major updates since Gemini 3 Pro launched in November 2025 and received an iterative update with Gemini 3.1 Pro in February 2026. While competitors pushed several flagship iterations into production—including OpenAI's GPT 5.5 Pro, GPT 5.6 Astro, and GPT 6 Astro alongside several updates to Anthropic's Claude line—Google focused its summer releases on smaller, cost-efficient systems, notes Computerworld, such as Gemini 3.8 Flash, according to 9to5Google.
Addressing questions about the release cadence, Kavukcuoglu—who recently took over as head of the Google DeepMind business unit from founder Demis Hassabis—stated that Google aims to release an early post-training output "as soon as possible." Kavukcuoglu expressed confidence in the division's standing, indicating that Google hopes to deploy Gemini 4 "much earlier" than the end of 2026, according to 9to5Google.
Gemini 4 Pro Architecture and Capabilities
Technical details published by Geeky Gadgets indicate that Gemini 4 Pro is built on an architectural checkpoint dubbed "Barium B." This checkpoint is designed to optimize processing speed, precision, and detail.
According to reports highlighted by Universe of AI via Geeky Gadgets, key focus areas for Gemini 4 Pro include:
- Software Engineering: Optimized code generation, debugging, and complex problem-solving.
- 3D Rendering: Advanced capabilities aimed at generating complex designs with high accuracy and speed.
- Real-Time Data Analysis
- Antigravity Integration: Internal testing is already underway utilizing Gemini 4 to power Google's Antigravity project, per 9to5Google.
- Recursive Self-Improvement (RSI): DeepMind is exploring potential integration of RSI, an unconfirmed capability intended to allow the system to iteratively refine its internal algorithms and code over time.
Leaked benchmarking metrics reported by Universe of AI suggest that Gemini 4 Pro consistently outperforms rival systems, such as Opus 5.5 and GPT6 Astra, across specialized technical and creative workloads.
Release Timeline and Market Context
Google DeepMind is targeting an official launch for Gemini 4 Pro scheduled for October 2026, following delays introduced to further polish and refine model behavior, according to Geeky Gadgets.
Rather than framing the technological race purely around artificial general intelligence (AGI), Kavukcuoglu emphasized that DeepMind's current objective is centered on practical reliability: "The conversation is more about are we able to build intelligent agents that we can trust," he stated, as reported by 9to5Google. The pending release sets up renewed competition against emerging proprietary releases and open-source models such as GLM 5.5 Flash, Kimmy K4, and Deepseek 4.1 Pro.
Frequently asked questions
What happened to Google's Gemini 3.5 Pro?
Google decided not to release Gemini 3.5 Pro, instead pivoting its development efforts to focus directly on launching Gemini 4 as its next flagship AI model.
When will Gemini 4 be released?
Gemini 4 has entered post-training safety and refinement, with Google DeepMind targeting a release much earlier than the end of 2026, specifically scheduled for October 2026.
What is the Barium B checkpoint in Gemini 4 Pro?
Barium B is the underlying architectural checkpoint powering Gemini 4 Pro, designed to increase processing speed, precision, and performance across 3D rendering and software engineering tasks.
Sources
- Google says Gemini 4 release is coming ‘as soon as possible’9to5Google · Sep 24, 2026
- Gemini 4 Pro vs Opus 5.5: How the New Models Compare - Geeky Gadgetsgeeky-gadgets.com · Sep 27, 2026
- Google plans Gemini 4 release before year-endComputerworld · Sep 25, 2026
How this story was made: the newsroom picked it up from Google News, gathered the full text of the sources above, and drafted it with AI assistance. Every factual claim was then checked against those sources before publishing (29 claims checked). Illustrations marked as AI-generated are not photographs. Spotted an error? Tell us.
Published September 28, 2026 at 00:17 UTC
